Dataset index sequence_frames_in_order

[jh27kim]

Hi, thanks for your great work.

I just finished downloading the entire dataset, but it seems like there are more sequences (36K) than reported in the paper which is 18K.

Is there any way I can sanity check this ?

Here's output of examples/print_co3d_stats.py

[shapovalov]

Thanks for your interest. CO3Dv2 was augmented with more sequences, and there are now indeed 36k sequences in total.

[jh27kim]

Alright, must have missed it, thank you !

[jh27kim]

Hi, I just noted that when I call sequence_frames_in_order from pytorch3d.implicitron.dataset.dataset_base.DatasetBase it gives me ~101 index which is different from the number of frames I downloaded (~202).

Seems like frames have been randomly selected internally Is this a correct behavior ? if so, could you please shared how it has been sampled ?